Abstract
Using the Swedish military enlistment test, this paper estimates the return to schooling for individuals belonging to different parts of the ability distribution. It also attempts to predict whether an endogenous test score causes bias in the “ability-specific” returns to schooling that varies with the test score. A significant finding is that a higher score in the test is associated with a higher return to schooling, but that the positive association is diminishing in the test score. In general, the bias in the ability-specific returns to schooling does not seem to vary with the test score level.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
In Card (1999), the causal return to schooling literature is surveyed.
The methods extend the policy evaluation literature by allowing heterogenous treatment effects.
This is the first paper to actually use the test score result from the Swedish Military Enlistment Battery in a cross-sectional study. In Nordin (2005), the Swedish Military Enlistment Battery is further discussed.
There are only a small number of women who have taken the Military enlistment test, and this group of women can hardly be considered a representative sample.
Except for 19 years of schooling, all potential years of schooling between 9 and 20 are contained in our schooling variable.
The variable used in this paper for measuring cognitive ability is in fact the general intelligence factor, G. For more information about the G factor, and the other group factors, see Carroll (1993).
For more information about the “test score groups” and the separate test results, see the Appendix.
When using the Swedish Military Enlistment test we end up with nine test score groups.
We do not report these test score group indicator coefficients because these variables do not yield any relevant information, and they should merely be treated as controls.
Lacking an actual experience measure, we use the standard method of constructing experience, i.e., exp = age - 7 - years of schooling.
Because of a positive relationship between schooling and the probability of belonging to the earnings equation sample, the full return to schooling is higher than the ones estimated in this paper.
As Griliches (1977) points out, when controlling for ability using a test score, schooling might be negatively correlated with the wage equation residual.
Both the decision to study after compulsory education and whether to choose a vocational or a theoretical study programme could then potentially create test score differentials between individuals.
Having 10 years of schooling means that, in the majority of the cases, the individual is an early dropout from upper-secondary education.
Using indicator variables for each of the test score groups instead of the discrete and ordered test score variable does not change the result.
The new SUN 2000 schooling variable with better precision does partly explain our comparatively high estimate.
The appendix describes the family income measure and explains the construction of the variable.
Table 3 reports descriptive statistics for the educational attainment variables.
This model is not pursued any further because the estimated returns to schooling are conditioned on study programme and obtaining a degree, which means that it is a departure from the mainstream approach.
The difference in the ability-specific return to education between the first and the second and also between the first and the third test score groups is not statistically significant.
Another explanation for the “high” return for the lowest test score group is discussed in the next section.
By controlling for these variables, we have already shown that these nonlinearities do not cause the relationship between the return to schooling and the test score to change.
This is tested by changing the reference group.
An exception is the income premium for the 16th year of schooling for the third test score group, which is significant at the 10% level.
References
Angrist JD, Krueger AB (1991) Does compulsory school attendance affect schooling and earnings? Q J Econ 106(4):979–1014
Antelius J, Björklund A (2000) How reliable are register data for studies of the return on schooling? An examination of Swedish data. Scand J Educ Res 44(4):341–355
Becker GS (1975) Human capital, 2nd ed. University of Chicago Press, Chicago
Björklund A, Edin P-A, Fredriksson P, Krueger AB (2004) Education, equality and efficiency—an analysis of Swedish school reforms during the 1990s, IFAU report 2004: 1. Institute For Labour Market Policy Evaluations, Uppsala
Blackburn ML, Neumark D (1995) Are OLS estimates of the return to schooling biased downward? Another look. Rev Econ Stat 77(2):217–230
Card D (1999) The causal effect of education on earnings. In: Ashenfelter O, Card D (eds) Handbook of labor economics, vol 3. Elsevier, Amsterdam, pp 1801–1863
Carneiro P (2002) Heterogeneity in the returns to schooling: implications for policy evaluation, Ph.D. thesis. University of Chicago, Chicago
Carneiro P, Hansen KT, Heckman JJ (2001) Removing the veil of ignorance in assessing the distributional impacts of social policies. Swed Econ Policy Rev 8:273–301
Carneiro P, Hansen KT, Heckman JJ (2003) Estimating distributions of treatment effects with an application to the returns to schooling and measurement of the effects of uncertainty on schooling choice. Int Econ Rev 44(2):361–442
Carroll JB (1993) Human cognitive abilities. A survey of factor-analytic studies. University Press, Cambridge
Chesher A (2003) Identification in nonseparable models. Econometrica 71(5):1405–1441
Griliches Z (1977) Estimating the return to schooling: some econometric problems. Econometrica 45(1):1–22
Gould ED (2005) Inequality and ability. Labour Econ 12(2):169–189
Hansen KT, Heckman JJ, Mullen KJ (2004) The effect of schooling and ability on achievement test scores. J Econom 121(1–2):39–98
Ingram BF, Neumann GR (2006) The returns to skill. Labour Econ 13(1):35–59
Juhn C, Murphy KM, Pierce B (1993) Wage inequality and the rise in returns to skill. J Polit Econ 101(3):410–442
Kjellström C (1999) Omitted ability bias and the wage premium for schooling: new Swedish evidence. In: Essays on investments in human capital, Ph.D. thesis. Swedish Institute for Social Research, Stockholm University
Meghir C, Palme M (2003) Ability, parental background and education policy: empirical evidence from a social experiment. Working paper series no. W03/05, Institute for Fiscal Studies
Mincer J (1974) Schooling, experience and earnings. Colombia University Press, New York
Neal DA, Johnson WR (1996) The role of premarket factors in black–white wage differences. J Polit Econ 104(4):869–895
Nordin M (2005) Studies in educational attainment and returns to schooling, Lic. thesis. Lund University
Winship C, Korenman S (1997) Does staying in school make you smarter? The effect of education on IQ in the bell curve. In: Devlin B, Fienberg SE, Resnick DP, Roeder K (eds) Intelligence, genes and success: scientists respond to the bell curve. Copernicus Press, New York, pp 215–234
Acknowledgment
I am grateful for comments by the editor, two anonymous referees, Peter Fredriksson, Anna Meyer, Inga Persson, Dan-Olof Rooth, and Mårten Wallette. The data has been financed through a research grant from the Swedish Council for Working Life and Social Research (FAS) that is gratefully acknowledged. This study is part of the project “Ethnic discrimination and Swedish-specific human capital.”
Author information
Authors and Affiliations
Corresponding author
Additional information
Responsible editor: Christian Dustmann
Appendices
Appendix
The test score groups
The transformation of the results of the separate tests into the nine-point scale shows changes during the time period, but because our data contains the results of the separate tests, we are able to construct a test score measure that is time consistent. Information is missing for one, two, or three of the separate test score results for 2.7% of the observations. In these cases, we use the average of the other test score results as a proxy for the missing test score result. Excluding these observations from the sample does not change the estimates.
Using the sum of the separate test score results instead of G in the wage regressions does not change any of the results found in this paper. Dividing the test score groups into 10 deciles, based on either G or the sum of the separate test score results, the estimates change in the tails of the ability distribution because the test score groups in the tails then get larger. This means that we lose the opportunity to analyze what takes place at the ends of the tails, e.g., that there are underachievers in the lowest test score groups and that the relationship between the test score and the return to schooling differs for graduates compared to the rest of the sample.
Family income
Family income is computed in the following manner. The mother’s and father’s average earnings for the years 1970, 1975, and 1980 are first computed. All earnings are in the 1980 prices. If some of the earnings are zero, an average of the remaining earnings is computed. The mother’s and father’s average incomes are then added.
Rights and permissions
About this article
Cite this article
Nordin, M. Ability and rates of return to schooling—making use of the Swedish enlistment battery test. J Popul Econ 21, 703–717 (2008). https://doi.org/10.1007/s00148-006-0131-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00148-006-0131-6